Adolescent dating violence is alarmingly common and exacts a substantial toll on public health. It exhibits substantial stability and may become deeply entrenched by adulthood. Given these difficulties, the prevention of adolescent dating violence has become a concern of the CDC, NIH, and several noted investigators. We, in concert with several colleagues, offer the following set of observations, which frame the problem at hand and the research that we believe needs to be done: (1) Clearly we've only just begun to learn how to pre- vent dating violence. Rigorous efficacy trials have been rare and have produced sometimes promising but sometimes mixed results. We need to do better. (2) Dating violence prevention would be enhanced by the identification of malleable risk factors that can become new intervention targets. (3) There are clear indications in the research literature that adolescent couples' relationship dynamics are important factors in dating violence. However, the field is just beginning to understand what those dynamics are. (4) We believe it highly likely that the discovery of additional relationship dynamics that undergird dating violence will ultimately lead to enhanced interventions to prevent it. Accordingly, we are proposing an observational [longitudinal] study of 100 New York City 14- to 18-year- old dating couples in order to test the hypothesis that some adolescent couples may inadvertently provide one another with basic training in hostile behavior, ultimately leading to violence. This hypothesis stems from Gerald Patterson and colleagues' groundbreaking work showing the power of such coercive processes in explaining parent-child and sibling dynamics that contribute to child aggression. While some have hypothesized the role of coercive process in adolescent couples' dating violence, the model has yet to be directly tested. To evaluate the role of coercion in youth dating violence, we intend to apply our team's expertise in intimate violence, couples observation, and development to the execution of the following aims: Aim 1: Test the fundamental basic training premise of the coercion model - that the occurrence of aversive and non-aversive behaviors is tied to their effectiveness in terminating conflict. We hypothesize that behaviors more effective in terminating conflict will be performed at a higher rate. If aversie behavior is more effective than non-aversive behavior, it will be performed at a proportionally higher rate; and vice versa. ? Aim 2: Test the association of coercive process and dating violence. We hypothesize that violence will be elevated [and/or increasing] in couples for whom aversive behaviors are proportionally more effective than non-aversive ones in terminating conflict, and show higher rates of reciprocation during conflict. ? Aim 3: Determine whether other previously identified risk factors (e.g., anti-sociality, family violence) explain dating violence via their impact on coercion, or whether coercion explains [levels and/or change in] dating violence independent of these factors' influences.